@dingchilling🫪

@bryan_johnson @_katetolo tell them that 1% stat about kate. it would be a good ice breaker 🥰

I’m meeting Kate’s parents for the first time. Do you think they’ll like me? I’m flying 17 hrs. She’s from Australia and international travel increases aging. But I really love her so it’s worth the cost.

@pangramlabs lfg!🦊

10k followers! Thank you everyone 🦊🧡

@KeyTryer is there an emulator on ai studio for preview or you have to keep installing the package over and over?

I couldn't find a good pomodoro timer app that worked the way I wanted, and just asked Gemini to make one, it made an installable package

@iScienceLuvr i'd log off for a while and enjoy the payout. anyway don't let the hate get to you it's mostly elon dickriding bots.

Damn Elon insulted me now my notifications are filled with hundreds of people insulting me lol

they are making cdramas of dario🤣🤣🤣

@alexandrabotez 1) what

I’ll be mentoring this AI safety creator bootcamp! If you want to create content about AI and the future, you should apply. Free to attend + room and board covered. Applications are rolling, so apply sooner rather than later 😊

@drummatick "cause AI distribution is trained to mimic what humans do" probably true for base models but instruction tuned models are selected for specific chatbot style and easily detectable. just try pangram and you'll be surprised how wrong this is.

Reason why AI detectors will never work and the problem can be better scoped. Even if you’re non technical you’ll understand it. You can’t “detect” whether a piece of writing is AI generated or not unless you know what the true data distribution for that person looks like. Every person has their own way of writing and given that AI learned from all across the internet, it also has that distribution. An amalgamation of humanity. There’s no “true AI” only data distribution out there. You can’t draw a plane that separates humans from AI generated writing cause AI distribution is trained to mimic what humans do. The training guarantees that. What can be done is, for a person, look at their past writing patterns, map it to a distribution and then check whether a new article written by them comes from the same distribution or not. Then you’ve a higher chance of detecting any anomalies. That’s the best of what you can do. You can still not make a claim whether it’s AI or not but you can make a claim it is definitely not that persons writing

@seconds_0 now that I know this information you should delete this tweet

converts to multiple formats as well. never heard of this app before. was asking claude to resize and compress an image and it said we could do it using this app faster

just found out this google app that you can use to compress images right from your browser (all thanks to claude) squoosh.app

@NateWitkin @alexolegimas @TuhinChakr @pangramlabs not denying that an abnormal number of rare n-grams could indicate non-human writing. you just can't trace them back to a singular source as implied in the blog and it's not plagiarism if you can't trace all of them back to a singular source using them in all the same contexts.

I think there's a pretty strong case to be made that reusing 437 rare phrases in a single piece of writing is plagiarism. I don't see it as particularly important that some of them appear in multiple places, and that some are more generic than others. Quite a few read to me as unique turns of phrase, enough that I think it'd bother most people to find out a human writer had lifted them. I could be convinced the other way if it turns out that human-written pieces re-use an analogous number of rare n-grams, but I doubt they do.

This from @TuhinChakr is brilliant. That prize winning story from Granta? Turns out it's just a bunch of random whole phrases taken directly from existing text on the internet. Tool allows you to trace those n-grams directly to their source, which is mostly random fanfiction. tuhinchakrabarty.substack.com/p/ai-slop-gran…

@TuhinChakr @NateWitkin @alexolegimas @pangramlabs @KelseyTuoc the blog is confidently attributing the source and i am showing you the expressions are more common than you think. it doesn't matter they are all from fanfic or not-- they just can't be attributed to a single source. that's my only point of contention.

@TuhinChakr @NateWitkin @alexolegimas @pangramlabs @KelseyTuoc what? i don't follow. i was only opining on the plagiarism point raised by nate. you cannot steal words and expressions and you certainly cannot trace them back to a single source like the blog says it does.

@TuhinChakr @NateWitkin @alexolegimas @pangramlabs @KelseyTuoc i am sorry but the blog it says it was "stolen" from that particular fan fic page (which is now down by your own submission) i am showing you it is not rare and can be found on many pages still up so there is no guarantee that it was drawing from that page 🫪🫪🫪

@dingchilling @NateWitkin @alexolegimas @pangramlabs @KelseyTuoc Lmao thats all fanction in your screenshot 😂

@NateWitkin @alexolegimas @TuhinChakr @pangramlabs similarly "swallows a shout" can still be found on multiple pages on the internet @KelseyTuoc

@NateWitkin @alexolegimas @TuhinChakr @pangramlabs just a google search for the exact phrase "Something coiled inside her" gets so many hits both before and after the slopocalypse began.

looks like a tile from a persian monument

@AnthropicAI api or it didn't happen

Last month we launched Project Glasswing, our collaborative AI cybersecurity initiative. Since then, we and our partners have found more than ten thousand high- or critical-severity vulnerabilities in essential software.